Retrieval Practice: The Testing Effect and How to Use It

Retrieval practice is learning by pulling information out of memory, through quizzes, flashcards, or writing down what you remember, instead of putting it back in by rereading. The testing effect is the research finding behind it: retrieving information improves long-term retention more than spending the same time restudying it.

It is one of the most replicated results in learning science. Researchers have studied it for more than a century, in labs and in real classrooms, with learners from kindergarten to medical school.

This page is the science: the history, the landmark studies, the classroom evidence, the theories of why retrieval practice works, and where it breaks down. For a step-by-step student guide, see what active recall is and how to use it.

What is retrieval practice?

Retrieval practice is any activity where you produce information from memory and, ideally, check it afterward. Common retrieval practice examples:

• Answering a flashcard out loud before flipping it
• Writing everything you remember about a topic on a blank page (a brain dump)
• Doing past-paper or end-of-chapter questions without notes
• Explaining a concept aloud without looking
• A low-stakes quiz at the start or end of a class

Researchers call the strategy retrieval practice, practice testing, or test-enhanced learning, and the result the testing effect. Students usually say active recall.

The research separates two kinds of benefit. Direct effects come from the act of retrieving: the memory itself gets stronger. Mediated effects come from how a test changes what you do next; for example, studying after a test is more effective because the test showed you what you didn't know. A third, the forward testing effect, is covered below: testing old material can help you learn new material.

A short history of the testing effect

The idea is older than most study advice:

• 1906. Psychologist Edward Thorndike recommended "the active recall of a fact from within" over passively taking it in.
• 1909–1917. Early experiments, including Arthur Gates's work on "recitation" with schoolchildren, found that time spent reciting from memory improved learning.
• 1939. Herbert Spitzer tested 3,605 sixth graders from 91 elementary schools. Students who took a practice test after reading a short text did better on a test a day or a week later than students who didn't.
• 1989. Interest had faded so far that one paper was titled "The 'testing' phenomenon: Not gone but nearly forgotten."
• 2006. Henry Roediger and Jeffrey Karpicke showed the effect with prose passages and essay-style recall: rereading won on a test five minutes later, and recall won two days and a week later. That study, charted in our active recall guide, helped restart the field.

By 2013, a major review counted several hundred experiments over more than 100 years showing that practice testing improves learning and retention.

The landmark studies

Retrieval, not extra studying, does the work

In 2008, Karpicke and Roediger had college students learn 40 Swahili–English word pairs, such as mashua–boat, through cycles of study and test. Once a student recalled a pair, it was treated one of four ways: kept in both study and tests, kept in tests only, kept in study only, or dropped entirely. The last option is the usual advice: study something until you can recall it, then drop it.

Every group learned the list equally well, and every group predicted it would remember about half of it a week later. The real results were very different.

Repeated studying after the first correct recall added nothing measurable. Repeated testing produced a huge benefit: individual scores didn't even overlap (63–95% with repeated testing, 10–60% without). And students' predictions bore no relation to what they actually remembered.

One honest footnote: a 2016 replication by Soderstrom, Kerr, and Bjork confirmed the testing result, but found that restudying also helped once spacing between items was equalized. Repeated retrieval is powerful; restudying isn't worthless.

The gap grows with time

Testing helps most when the final test is far away. In a study by Runquist summarized in the 2013 review, practice-tested word pairs were recalled better than untested ones 10 minutes later (53% vs 36%). After a week, the gap was 35% vs 4%. In studies that compare testing with restudying, rereading often matches or beats testing minutes after studying. The difference shows up days later, which is why people underrate retrieval. More on that in active recall vs passive learning.

Testing helps you learn the next thing too

The forward testing effect is the finding that recalling earlier material improves learning of new material that comes after it. Reviews describe it as well replicated in the lab, and it has a practical form.

In a 2013 Harvard study by Karl Szpunar and colleagues, students watched a 21-minute statistics lecture in four segments. One group took a short quiz after every segment; others restudied or did arithmetic between segments. Everyone was tested on the fourth segment. The quizzed group:

• mind-wandered during 19% of random checks, versus 39–41% for the other groups
• took notes on 24% of slides, versus 7–9%
• scored 89% on the fourth segment, versus 65–70%
• reported less anxiety about the final test

Earlier lab work found that short tests between study blocks protect against interference from earlier material, which otherwise builds up over a long session.

It transfers beyond the exact question

Does recall help with questions you haven't practiced? Often. In one study summarized in the 2013 review, students who practiced short-answer questions with feedback did better a week later on new inference questions than students who restudied, including questions that applied ideas about bat wings to aircraft design.

A 2018 meta-analysis by Steven Pan and Timothy Rickard pooled 192 transfer results from 122 experiments (10,382 participants) and found a moderate average benefit (d = 0.40). Transfer was strongest to new question formats, to application and inference questions, and to medical diagnosis problems. It was weakest to untested material seen during study. After correcting for publication bias, the authors found transfer often vanished unless certain conditions were met, such as practice that required the same kind of response as the later test.

Does retrieval practice work in the classroom?

Yes. The strongest evidence is a 2021 systematic review by Pooja Agarwal, Ludmila Nunes, and Janell Blunt. They screened nearly 2,000 abstracts and kept only studies run in real schools and classes: 50 experiments, 49 effect sizes, and 5,374 students. 57% of the effects were medium or large. Benefits held across grade levels, subjects, quiz formats, and delays. One gap: only 6% of experiments were run outside Western, educated, industrialized, rich, democratic countries.

A typical example is a 2011 series of experiments in a sixth-grade social studies class. Students took three short, low-stakes multiple-choice quizzes on some of the course material. On their actual chapter and semester exams, they scored better on quizzed material than on material that wasn't quizzed or was simply presented twice.

The 2013 review lists testing effects in preschoolers, school students, medical students, and older adults, with benefits lasting months in some classroom studies. Low-stakes quizzing also doesn't seem to raise anxiety: a 2014 classroom study reported that retrieval practice programs reduced middle and high school students' test anxiety. For more on that, see managing test anxiety.

Why does retrieval practice work?

The effect is clearer than its explanation. The main theories:

• Elaborative retrieval. Searching memory activates related information, which gets stored along with the answer and creates more routes back to it. In one study, people who practiced recalling child from mother later recalled it better from a new cue, father, than people who restudied the pair.
• Better mediators. When you learn a foreign word through a keyword or image, retrieval practice makes those links stronger. In a Swahili vocabulary study, tested learners were better at remembering their own keyword links a week later.
• Episodic context. Karpicke's group argues that each retrieval updates the context stored with a memory, so future searches start in the right place. Experiments have separated this from simple elaboration, and the 2014 meta-analysis found only limited support for elaboration accounts.
• Bifurcation. Testing splits items into two groups: ones you retrieve get a big boost; ones you fail on, without feedback, get none. That explains why testing can lose on a test minutes later and why feedback matters so much.

Indirect benefits add to these: tests show you what you don't know, and repeated recall seems to help organize what you've learned. No single theory explains every result yet, and several mechanisms probably contribute.

When does retrieval practice work best?

Rowland's 2014 meta-analysis compared testing with restudying across 159 effect sizes, mostly from lab studies. The average benefit was moderate (g = 0.50), but it varied a lot with conditions.

The main moderators, from that analysis and the 2013 review:

• Feedback. Studies with feedback averaged g = 0.73, versus 0.39 without.
• Retrieval success. Without feedback, practice you mostly fail produced no reliable benefit. Practice you mostly got right produced a solid one.
• Delay. The effect was larger when the final test came a day or more later (0.69) than within the same day (0.41).
• Format. Practice that makes you produce an answer tends to beat recognition. Among studies where learners got feedback or recalled most items, Rowland found average benefits of 0.72 for cued recall and 0.81 for free recall, versus 0.36 for recognition tests such as multiple choice. Success matters too: in four experiments, Smith and Karpicke (2014) found well-built multiple-choice questions helped about as much as short-answer ones, except when short-answer practice was made easier to get right (more in our practice quiz guide).
• Dosage and spacing. Recalling an item several times beats once, with diminishing returns. Repeated tests back to back add little; tests spread over days add much more. That is where spaced repetition comes in.

Limits and open questions

Retrieval practice is robust, not magic. The caveats:

• Short delays. On a test minutes later, rereading can match or beat it.
• Material you can't retrieve. Without feedback, failed attempts teach little. Pretesting is the exception: guessing before you study helps, because the answers come right after (see the pretesting effect).
• Complex material. In 2015, van Gog and Sweller argued the testing effect shrinks as materials get more complex; Karpicke and Aue replied that it holds with complex materials. Evidence on inference and transfer questions suggests it helps, but less reliably than with facts.
• Who was studied. Most lab studies use college students, and the classroom evidence comes mainly from Western countries.
• Learners don't believe it. Across several studies, people predict that restudying will work as well or better. Judge it by your results a week later.

What the research means for how you study

To put this into practice, pick a method from the active recall guide, which ends with a 14-day starter plan, and time your reviews with the forgetting curve explained.